[PATCH 2/4 V2] sparc: Add pci32_unmap_page.

[PATCH 2/4 V2] sparc: Add pci32_unmap_page.

[Kristoffer Glembo]

Signed-off-by: Kristoffer Glembo <kristoffer@gaisler.com>
---
 arch/sparc/kernel/ioport.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)

diff --git a/arch/sparc/kernel/ioport.c b/arch/sparc/kernel/ioport.c
index cea49c0..1b9f07c 100644
--- a/arch/sparc/kernel/ioport.c
+++ b/arch/sparc/kernel/ioport.c
@@ -524,6 +524,13 @@ static dma_addr_t pci32_map_page(struct device *dev, struct page *page,
 	return page_to_phys(page) + offset;
 }
 
+static void pci32_unmap_page(struct device *dev, dma_addr_t ba, size_t size,
+			     enum dma_data_direction dir, struct dma_attrs *attrs)
+{
+	if (dir != PCI_DMA_TODEVICE)
+		mmu_inval_dma_area((unsigned long)phys_to_virt(ba), PAGE_ALIGN(size));
+}
+
 /* Map a set of buffers described by scatterlist in streaming
  * mode for DMA.  This is the scather-gather version of the
  * above pci_map_single interface.  Here the scatter gather list
@@ -646,6 +653,7 @@ struct dma_map_ops pci32_dma_ops = {
 	.alloc_coherent		= pci32_alloc_coherent,
 	.free_coherent		= pci32_free_coherent,
 	.map_page		= pci32_map_page,
+	.unmap_page		= pci32_unmap_page,
 	.map_sg			= pci32_map_sg,
 	.unmap_sg		= pci32_unmap_sg,
 	.sync_single_for_cpu	= pci32_sync_single_for_cpu,
-- 
1.6.4.1

Re: [PATCH 2/4 V2] sparc: Add pci32_unmap_page.

[Sam Ravnborg]

On Thu, Jan 13, 2011 at 10:05:33AM +0100, Kristoffer Glembo wrote:
> Signed-off-by: Kristoffer Glembo <kristoffer@gaisler.com>

I totally miss _why_ you add this function.
Just a small comment in the changelog would help.
It may due to my ignorance of all the dma stuff...


> ---
>  arch/sparc/kernel/ioport.c |    8 ++++++++
>  1 files changed, 8 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/sparc/kernel/ioport.c b/arch/sparc/kernel/ioport.c
> index cea49c0..1b9f07c 100644
> --- a/arch/sparc/kernel/ioport.c
> +++ b/arch/sparc/kernel/ioport.c
> @@ -524,6 +524,13 @@ static dma_addr_t pci32_map_page(struct device *dev, struct page *page,
>  	return page_to_phys(page) + offset;
>  }
>  
> +static void pci32_unmap_page(struct device *dev, dma_addr_t ba, size_t size,
> +			     enum dma_data_direction dir, struct dma_attrs *attrs)
> +{
> +	if (dir != PCI_DMA_TODEVICE)
> +		mmu_inval_dma_area((unsigned long)phys_to_virt(ba), PAGE_ALIGN(size));
> +}

How about a patch that fix it so mmu_inval_dma_area() takes a void *
as first argument?
In most cases I see a cast from void * => unsigned long where we call
mmu_inval_dma_area().

In sbus_alloc_coherent() we use the result of __get_free_pages()
but the general pattern is that the result of __get_free_pages()
is cast to the right type when used.

So if we did:

void * va;

va = (void *)__get_free_pages(GFP_KERNEL|__GFP_COMP, order);
Then we could remove all the other casts.


	Sam

Re: [PATCH 2/4 V2] sparc: Add pci32_unmap_page.

[Kristoffer Glembo]

Hi Sam,

Sorry for my sluggish replies.

Sam Ravnborg wrote:
> On Thu, Jan 13, 2011 at 10:05:33AM +0100, Kristoffer Glembo wrote:
>> Signed-off-by: Kristoffer Glembo <kristoffer@gaisler.com>
> 
> I totally miss _why_ you add this function.
> Just a small comment in the changelog would help.
> It may due to my ignorance of all the dma stuff...
> 


It is part of the DMA API and needed by our ethernet driver. I guess no PCI driver (on sparc) used it before.


> 
> How about a patch that fix it so mmu_inval_dma_area() takes a void *
> as first argument?
> In most cases I see a cast from void * => unsigned long where we call
> mmu_inval_dma_area().
> 


Ok.


/Kristoffer